Key points are not available for this paper at this time.
Although numerical models are essential to hurricane forecasting, many other applications require only statistical depiction of the wind distribution. In Holland’s 1980 parametric profile, radius of maximum wind, maximum wind, and a measure of the profile width describe the radial variation of the axisymmetric wind. Variants of the Holland profile are used to predict insurance underwriting risk, ocean response, and storm-surge inundation. Since these calculations guide high-stakes financial and emergency managment decisions, it is logical to test them against observations. The Hurricane Research Division’s flight-level database archives observations were obtained by NOAA and U.S. Air Force Reserve aircraft. The data considered here are winds and geopotential heights observed during 606 lower- and midtropospheric flights into Atlantic and eastern Pacific tropical cyclones during 1977‐2000. The 493 profiles that meet quality control criteria are seasonally geographically representative. Least squares fits of the Holland model to these data provide evaluation of the parameters’ distributions and critical examination of the profile’s realism. Individual fitted profiles differ from the observations in a consistent pattern. The areas of strong winds in the eyewall and of nearly calm winds at the vortex center are too wide. Beyond 2 or 3 times the eye radius, the wind decreases too rapidly with distance from the center. Although the average bias in fitted profiles is, 1ms 21, the root-mean-square error is 4.2 m s 21 (5.2 m s21 for independent data). Maximum winds estimated from the fitted Holland-profile height‐wind relation average 2.5 m s 21 too strong with an rms error of 6.5 m s 21. The pattern of too strong wind spread over too much real estate exaggerates the occurrence of winds stronger than 50 m s 21 by ;50%.
Willoughby et al. (Wed,) studied this question.